I have worked with both DCOS and Kubernetes - the latter is way more coordinated and polished, with also better fundamentals underneath. This results in a situ

I have worked with both DCOS and Kubernetes - the latter is way more coordinated and polished, with also better fundamentals underneath. This results in a situation where a lot of different places have logical, reasonable reasons to pick k8s because it matches reasonably well 80% of their needs, and while everyone's 80% is different, there's enough overlap. This pulls in, like frame dragging, others who might not actually need it, but it's not because some scaling memes. In practice, a lot of initial intake for k8s was not bananas scaling arguments, but it actually solving problems for people who either had growing complex messes, or considered the spend associated with "good practices" that are often thrown as argument against kubernetes, like running separate cloud instances for every application. Similarly, everytime I look at Nomad and related stack, it turns out that various bits are missing, and DIY will be harder than with k8s-like stack because it's not properly designed to be extensible.